India’s AI Crossroads, Between Protectionism and Preparedness
The Ghost of Industrial Revolutions Past Haunts India’s Digital Future
As global technocrats, policymakers, and industry leaders gather at the India AI Impact Summit in New Delhi, the proceedings carry an undercurrent of urgency that extends far beyond the usual conference-circuit platitudes. This is not merely another gathering to celebrate technological progress. It is, in essence, a reckoning—a moment when India must confront the hard questions about its place in an artificially intelligent world.
Vivan Sharan and Vedika Pandey, technology policy experts who serve as the secretariat for the AI Knowledge Consortium, have articulated what might be called the three tests for India’s AI future: human capital readiness, access to enabling infrastructure, and institutional capacity. These are not abstract categories. They are the fault lines along which India’s technological trajectory will either succeed or falter.
The recent Economic Survey’s cautionary note—that rapid AI deployment could outpace the economy’s structural ability to reabsorb labour—deserves more attention than it has received. With over seven per cent of India’s GDP tied to the IT and IT-enabled services sector, the stakes are existential. Unlike previous industrial transformations that automated manual labour, the job displacement risks associated with AI are fundamentally different. They target the very cognitive skills that have been India’s export advantage for three decades.
The Compression of the Skills Ladder
Consider what is already happening. AI systems are increasingly capable of writing and improving their own code. This is not science fiction; it is the mundane reality of platforms like GitHub Copilot and OpenAI’s Code Interpreter. The ladder of skills that once trained generations of Indian programmers—from learning syntax to debugging to architectural design—is being compressed from below and above simultaneously.
For a country that built its global reputation on human-led services, this raises questions that cannot be answered with slogans about a demographic dividend. What happens when the entry-level coding jobs that absorbed millions of engineering graduates begin to evaporate? What happens when the comparative advantage in labour-intensive software development is eroded by machines that work for the marginal cost of electricity?
The temptation, as Sharan and Pandey rightly note, is protectionism through delay. If we slow down AI diffusion, the argument goes, we can save jobs. We can give our workforce time to adapt. We can shield our domestic industry from the gale of creative destruction.
This temptation is as old as industrial capitalism itself. And it is equally doomed to fail.
The British Textile Parallel and Its Lessons
The authors invoke a telling historical parallel: Britain’s attempt in the eighteenth century to protect its domestic textile industry by banning the export of advanced machinery. The logic was impeccable from the perspective of British industrialists. Keep the technology at home, maintain the competitive edge, and ensure that continental competitors cannot replicate what Manchester had achieved.
But history records the outcome differently. British machines were smuggled out nonetheless—disassembled, packed in crates mislabelled as agricultural equipment, and reassembled in mills across continental Europe. Britain could not stop the rise of competing manufacturing hubs. The attempt to hoard technology only ensured that when competition finally arrived, it came on terms that Britain could not control.
The parallel with contemporary India is instructive but not exact. Today’s protectionism would not target the export of machinery but the import of AI services, the deployment of AI tools, and the diffusion of AI capabilities across the economy. The logic, however, is identical: limit the spread of technology to preserve domestic advantage.
Yet the same forces that defeated Britain’s mercantilist ambitions are amplified in the digital age. Code moves at the speed of light. AI models can be accessed through virtual private networks, hosted on servers in jurisdictions with lighter regulation, or simply downloaded as open-source weights that no customs inspection can intercept. Slowing AI diffusion in India will not slow AI adoption by Indian companies’ global competitors. It will simply handicap Indian firms in a race they cannot opt out of.
The way forward, as Sharan and Pandey argue, is not to protect workers from AI but to invest in the capacity to assist them through transition. This means imagining a meaningful social safety net appropriate for the AI era—a concept that remains woefully underdeveloped in Indian policy discourse. It also means encouraging companies to invest in innovation and research rather than rent-seeking behind protective barriers.
The False Choice Between Cloud and Device
A second theme that emerges from the analysis is the false choice that often dominates debates on AI architecture: centralised computing exemplified by cloud services versus decentralised on-device computing. In reality, developing economies like India will likely operate somewhere in between, balancing brute force with resilience and access.
This is not merely a technical question. It is a question of political economy. Cloud computing, once assumed to be infinitely scalable, now faces profitability pressures as AI workloads become vastly more expensive. The computational requirements of training and running large language models are staggering. Every query to an AI system consumes energy and processing power that, in the cloud model, must be paid for by someone.
The orchestration of AI workloads between the cloud and last-mile devices—particularly smartphones, which are ubiquitous in India—represents a feasible alternative in many use cases. A farmer querying an AI system about crop prices does not need a billion-parameter model running on a remote server. A small trader seeking translation services does not need the full capabilities of GPT-4. Much can be done on the device itself, preserving privacy, reducing latency, and lowering costs.
This is also where sector-specific deployment becomes crucial. AI is not a single technology but a suite of capabilities that must be tailored to specific domains. Hospitals need AI that understands medical records and diagnostic protocols. Banks need AI that can detect fraud while complying with financial regulations. Courts need AI that can assist with case management without introducing bias. Factories need AI that can optimise supply chains and predict equipment failures.
Building such systems requires in-house expertise that understands both the technology and the domain in which it operates. In the past, organisations relied on IT departments to choose software, troubleshoot problems, and train staff on tools. AI is different. It is not a tool to be purchased and installed but a capability to be cultivated and integrated. Sharan and Pandey put it succinctly: building such enlightened workforces may be as important as building data centres.
The Institutional Capacity Gap
The third test—institutional capacity—may be the most difficult of all. India’s regulatory apparatus was designed for an analogue world. Its courts move at the pace of paper. Its legislatures debate in terms drawn from nineteenth-century political philosophy. Its executive branches operate through hierarchies and circulars that assume a stable, predictable environment.
AI does not respect any of this. It moves fast, breaks things, and creates externalities that existing institutional frameworks are ill-equipped to handle.
Sharan and Pandey draw a telling parallel with the regulation of encrypted messaging and digital assets. In both cases, legitimate concerns over illicit activity—terrorist communication, money laundering, tax evasion—have sometimes led to demands for draconian solutions. Ban encryption altogether. Criminalise cryptocurrencies. Shut down the platforms.
This is characteristic of a policy environment where the absence of precise investigative tools leads to “all-or-nothing” approaches. When regulators cannot easily distinguish between legitimate and harmful activity, the default response tends towards precautionary restrictions. The unintended consequences are real: the offshoring of digital asset entrepreneurs and innovation from India to the UAE and Singapore is not a theoretical risk but a documented outcome.
With AI, the stakes are even higher. Consider the range of threats that AI enables: sophisticated disinformation campaigns that can sway elections, deepfakes that can destroy reputations, automated cyberattacks that can cripple infrastructure, algorithmic bias that can entrench discrimination. Each of these requires a regulatory response that is technically nuanced and contextually appropriate. Yet the institutional capacity to craft such responses is precisely what India lacks.
If we do not build this capacity—if regulators, courts, and even the executive branch cannot bridge their technical knowledge gaps—we risk a future where online trust is managed through restrictions that constrain innovation, speech, creativity, and the future of work itself. The choice between vulnerability and regressive rulemaking is no choice at all.
The AI Knowledge Consortium Initiative
It is in this overarching context that the AI Knowledge Consortium, which consists of sixteen research-led institutions, and The Pioneer are convening a pan-India session on February 19th. The gathering brings together senior technology and policy leaders for a conversation on how AI is reshaping economies, institutions, and societies.
The framing question is precisely the right one: why do some economies move from experimentation to widespread use of new technologies while others do not? The answer, as economic historians have long shown, is not simply about the existence of the technology. It is about societal preparation—through skills, infrastructure, institutions, and systems that make technology inclusive.
India has advantages that should not be discounted. A large pool of technical talent, a vibrant startup ecosystem, a democratic polity that can accommodate diverse interests, and a regulatory system that, for all its flaws, is capable of learning and adaptation. But advantages are not guarantees. They must be mobilised, directed, and sustained through deliberate policy choices.
The Financial Dimension of AI Infrastructure
The financial aspect of this discussion deserves more attention than it typically receives. Cloud computing, once assumed to be infinitely scalable, now faces profitability pressures as AI workloads become vastly more expensive. The economics of AI are fundamentally different from the economics of previous digital technologies.
Training a single large language model can cost tens of millions of dollars. Running inference at scale requires data centre investments that strain even the balance sheets of hyperscalers like Amazon, Google, and Microsoft. For a developing economy like India, the capital requirements of building indigenous AI capabilities are daunting.
This is where the orchestration of workloads between cloud and device becomes not just a technical optimisation but an economic necessity. By pushing appropriate tasks to edge devices, India can reduce its dependence on expensive cloud infrastructure while still benefiting from AI capabilities. This is also where public investment in foundational models—trained on Indian languages, Indian data, and Indian contexts—can complement private sector innovation.
The danger is that India will repeat the pattern of previous technological transitions: waiting for the rest of the world to develop the technology, then importing it at high cost, then struggling to adapt it to local conditions. The countries that succeed in the AI era will be those that participate actively in its development, not merely its consumption.
The Labour Question Revisited
Underlying all of these considerations is the labour question. The Economic Survey’s warning about rapid AI deployment out pacing structural reabsorption capacity is not an argument against AI. It is an argument for preparation.
What would preparation look like? It would begin with a honest assessment of which jobs are most at risk. The conventional wisdom that AI threatens only routine cognitive work is already outdated. Creative professions, professional services, and even some forms of artisanal expertise are being augmented or replaced by AI systems. The boundary between what humans do and what machines do is shifting faster than any previous technological transition.
Preparation would also involve reimagining education and training. The model of front-loaded learning—school, then college, then a lifetime of work—is obsolete. Continuous learning, micro-credentials, and modular skill acquisition must become the norm. This requires institutional innovation that Indian education has been slow to embrace.
Finally, preparation would require a social safety net that acknowledges the reality of job transitions. Unemployment insurance, portable benefits, and support for entrepreneurship are not socialist luxuries; they are capitalist necessities in an era of rapid technological change. Workers who know they will be supported through transitions are more likely to embrace new technologies, not resist them.
Conclusion: The Tests Ahead
India stands at a crossroads. The path of protectionism—slowing AI diffusion, shielding domestic industry, preserving jobs through regulation—leads to stagnation and decline. It is the path that Britain took with textile machinery, that the United States took with steel in the 1970s, that every declining economy has taken when faced with technological competition. It does not work.
The path of open adoption—embracing AI, investing in transition, building institutional capacity—is harder but necessary. It requires facing uncomfortable truths about the durability of India’s talent advantage. It requires investments in infrastructure and skills that do not yield immediate political dividends. It requires regulatory capacity that India does not yet possess.
The three tests that Sharan and Pandey identify are not sequential. They must be addressed simultaneously. Human capital readiness, enabling infrastructure, and institutional capacity are mutually reinforcing. Fail on any one, and the entire project falters.
The AI Knowledge Consortium’s initiative on February 19th is a step in the right direction. Bringing together technology and policy leaders for serious conversation is essential. But conversation must lead to action. India’s AI future will not be determined by summits and white papers. It will be determined by whether the country can pass the three tests that matter.
The ghost of industrial revolutions past haunts every technological transition. The countries that succeed are those that learn the right lessons from history: that technology cannot be stopped, that workers must be supported, that institutions must adapt, and that the future belongs to those who prepare for it.
India has the talent, the demography, and the democratic institutions to succeed. Whether it has the will remains to be seen.
Q&A: Unpacking India’s AI Challenge
Q1: What are the “three tests” for India’s AI future identified by Sharan and Pandey?
A: The three tests are human capital readiness, access to enabling infrastructure, and institutional capacity. Human capital readiness refers to the workforce’s ability to adapt to AI-driven changes in skills and employment. Enabling infrastructure encompasses both physical infrastructure like data centres and computational resources as well as the orchestration between cloud and edge devices. Institutional capacity refers to the ability of regulators, courts, and government agencies to understand and respond to AI-driven challenges with technical nuance rather than resorting to blunt, regressive restrictions.
Q2: Why is protectionism through delay considered a “dead end” for India’s AI strategy?
A: Protectionism through delay—slowing AI diffusion to save jobs—is considered futile for several reasons. First, historical precedent shows that technology cannot be contained; like British textile machinery smuggled to Europe despite export bans, AI capabilities will find their way across borders. Second, in the digital age, code moves at the speed of light and can be accessed through VPNs, offshore servers, or open-source downloads that no regulation can fully block. Third, slowing AI adoption in India only handicaps domestic firms against global competitors who will adopt the technology regardless, leaving Indian industries vulnerable to more agile international players who route around or operate beyond such barriers.
Q3: What does the British textile machinery ban of the eighteenth century teach us about contemporary AI policy?
A: The British attempt to ban the export of advanced textile machinery to preserve its competitive edge offers a cautionary tale. Despite the restrictions, British machines were smuggled out and used to establish mills across continental Europe. Britain could not stop the rise of competing manufacturing hubs. The lesson is that trying to hoard technology or limit its spread is ultimately self-defeating. The resources spent on enforcement and the inefficiency subsidised by protection do not prevent competition; they merely delay adaptation and leave domestic industries unprepared when competition inevitably arrives. For India, this means investing in worker transition and innovation rather than trying to build walls against AI.
Q4: How does the regulation of encrypted messaging and digital assets foreshadow challenges in AI governance?
A: The regulatory response to encrypted messaging and digital assets reveals a pattern that will likely repeat with AI. When legitimate concerns over illicit activity—terrorist communication, money laundering, tax evasion—meet a lack of precise investigative tools, the default response tends towards “all-or-nothing” approaches: demands to ban encryption entirely or criminalise cryptocurrencies. This regulatory insecurity stems from gaps in supervisory capacity rather than a desire for censorship. The unintended consequences include driving innovation and entrepreneurs offshore, as happened when Indian digital asset businesses moved to the UAE and Singapore. With AI, without building institutional capacity for nuanced understanding, we risk similar outcomes—restrictions that constrain innovation, speech, and creativity while failing to address the underlying concerns.
Q5: What is the significance of the AI Knowledge Consortium’s February 19th session, and what question will it examine?
A: The AI Knowledge Consortium, comprising sixteen research-led institutions, is partnering with The Pioneer for a pan-India session on February 19th that brings together senior technology and policy leaders. The session will examine why some economies successfully move from experimentation to widespread use of new technologies while others do not. This question gets to the heart of India’s AI challenge: technological capability alone is insufficient. Success depends on societal preparation through skills, infrastructure, institutions, and systems that make technology inclusive. The gathering represents an attempt to bridge the gap between technical expertise and policy formulation, recognising that India’s AI future will be shaped by how effectively these two domains can communicate and collaborate.
